White House Sets 30-Day Security Review for Frontier AI Models
2 min readThe federal government is about to get a formal look at the most powerful AI models before the public does. The White House is finalizing a voluntary framework with OpenAI, Anthropic, and Google that would give federal agencies up to 30 days to review new frontier AI models for national security risks before release. An announcement is expected before August 1.
A 30-day pre-release gate
Under the plan, a company preparing to launch a “covered frontier model” would submit it for a pre-release security evaluation. Agencies would then have a 30-day window to test the system for cybersecurity weaknesses and other national security concerns. The review is voluntary on paper, but with the three largest labs signed on, it effectively sets a threshold that rivals will be expected to clear.
Rooted in a June executive order
The framework rests on Executive Order 14409, “Promoting Advanced Artificial Intelligence Innovation and Security,” which took effect on June 2, 2026. The order directed the National Security Agency, the Cybersecurity and Infrastructure Security Agency, the Treasury Department, and the White House’s national cyber director to build a classified benchmarking process. That process defines which systems count as covered models and sets out how they are reviewed before release.
Which frontier AI models are covered
Officials have told the companies that the evaluations will be handled by the Commerce Department’s Center for AI Standards and Innovation and the NSA. The White House sent a draft of the framework to Anthropic, OpenAI, and Google about two weeks ago, and the three firms suggested edits. That points to an unusual dynamic: the companies whose models are being reviewed are helping shape the rules of the review.
Why it matters
A 30-day gate would be the most concrete US step yet toward treating frontier models like sensitive dual-use technology. Supporters argue that a short, security-focused review is a reasonable safeguard as models grow more capable at coding, cyber operations, and scientific research. Critics counter that a voluntary process shaped by the biggest labs could harden their lead, since smaller developers may lack the resources to navigate a federal review. Because the framework is voluntary, its real force will depend on whether the government later ties it to procurement or contracts.
Watch for the official announcement in the coming days, along with the fine print on what triggers a review and whether the 30-day clock can be extended.